On Mon, Aug 31, 2026 at 07:38:21PM +0200, Marek Vasut wrote:
> On 8/31/26 7:13 PM, Conor Dooley wrote:
> 
> Hello Conor,
> 
> > > > > > >         then:
> > > > > > >           required:
> > > > > > >             - "#clock-cells"
> > > > > > > +        - "#address-cells"
> > > > > > > +        - "#size-cells"
> > > > > > 
> > > > > > else, false? But I'm also not sure what the benefit is of requiring
> > > > > > addr/size cells when the child node is not?
> > > > > The subnodes of this syscfg node have reg property , see [PATCH v4 0/4]
> > > > > arm64: dts: st: Add STM32MP2 USB support .
> > > > 
> > > > Right, but I was questioning why the properties describing the reg
> > > > property in children is required when the children (and thus the reg
> > > > property in them) were not made required.
> > > 
> > > Do you mean this patternProperties should be made required in here ?
> > > 
> > > "
> > > +patternProperties:
> > > +  "^phy@[0-9a-f]+$":
> > > "
> > > 
> > > Could you please give me a hint how to do that ? git grep gives me nothing.
> > 
> > I mean, if the phys aren't going to be required I would not make the
> > cells properties required. I seem to recall a warning that dtbs_check
> > can emit when there are size/addr cells properties in a node where there
> > are no children, which is why I questioned this in the first place.
> > 
> > Otherwise, if the phys are to be made required (no idea if they should
> > be), I can only recall this being done with the non-pattern variant. IOW
> > required:
> >    - phy@0
> >    - phy@1
> The PHYs are part of the hardware and are (going to be) described in the SoC
> DT, so I would argue they should be described as required ?

Yeah, and there's almost certainly no chance of them being required
being problematic is there, since they don't impact existing code, but
have a new (or at least a previously unused on this platform) driver.